Google DeepMind adds Cheminformatics Skills to Antigravity for drug-discovery workflows
On 3 Nov 2026 Google DeepMind published Cheminformatics Skills for Google Antigravity, turning the product into a collaborative science workbench. In an official lung-cancer case study, an agent searches thousands of compounds, runs matched molecular pair analysis, and proposes rigidified macrocycles with synthetic routes in about an hour; the post says a similar iteration often took multiple days. The skills build on open-source RDKit with connectors to ChEMBL and PubChem. They are available today via the Antigravity app’s Science plugin; enterprise users can sign in with Google Cloud. DeepMind is offering up to 10,000 scientists the highest Antigravity usage tier free for six months (rolling onboarding, subject to eligibility, capacity, and selection); applications are open. The post also states that sparse data and assay noise limit predictive precision, and that computational designs remain hypotheses that need experimental validation.
Verified against Google DeepMind’s official post Agentic cheminformatics with Google Antigravity (November 3, 2026, Science): DeepMind introduces Cheminformatics Skills for Google Antigravity, aimed at cutting time chemists spend wrangling multi-source molecular data and turning the surface into a collaborative science workbench.
Drug design, the post says, typically means curating complex data from charts, figures and literature, training predictive models, and proposing candidates — often multiple days per iteration. Using real-world lung-cancer treatments (including the tyrosine kinase inhibitors crizotinib and lorlatinib), an agent searches thousands of compounds, filters irrelevant molecules, documents sources, runs matched molecular pair analysis, builds potency models, and proposes new rigidified macrocycles with synthetic routes. DeepMind states this end-to-end walkthrough takes about an hour.
The skills split labour: the language model handles workflow orchestration, database navigation and scientific reasoning; deterministic chemistry tools run molecular calculations. They are grounded in open-source RDKit, with connectors to curated ChEMBL and PubChem resources. An interactive viewer lets scientists browse compounds, inspect metadata and plot relationships (for example potency versus rotatable bonds). The agent can also produce electron-pushing diagrams and export results into common chemistry software formats.
Availability (per the post): available on Google Antigravity today — download the app and install the Science plugin; enterprise users can choose “Use Business Account” and sign in with Google Cloud credentials. To advance research pace, DeepMind is offering up to 10,000 scientists the highest usage tier in Antigravity free for 6 months (onboarded on a rolling basis, subject to eligibility, capacity, and selection). Applications are open.
Limits (per the post): sparse data, inter-lab assay noise, and narrow dynamic ranges limit predictive precision even for strong models; computational designs are testable hypotheses, not finished drugs. Every proposed design must still be synthesized and validated experimentally.
Figures, timing claims, offer terms and caveats above are from DeepMind’s original post, not independent IdeaDawn measurements.